Incident response lifecycle

Incident response is a cycle that starts before the incident

Preparationplans, team, tools,training, exercisesDetectionalerts and reports:something is wrongAnalysisis it real? scope,severity, priorityContainmentstop the spread,keep the evidenceEradicationremove the attackerand the way inRecoveryrestore, verify,watch closelyLessons learnedwhat to change,and who owns itevery incident improves the preparation for the nextNIST SP 800-61 Rev. 3 (2025) maps the same work onto the CSF 2.0 functionsGV · ID · PRDetect (DE)Respond (RS)Recover (RC)Improve (ID.IM)Incident response follows a cycle that starts long before the incident and ends by improving the next one.Most of the outcome is decided in preparation: a plan, a trained team, logs that exist, backups that restore.The 2025 revision of NIST’s guide folds response into the Cybersecurity Framework: Govern, Identify and Protectprepare; Detect finds; Respond contains and eradicates; Recover restores; improvement runs throughout.

The incident response lifecycle: preparation, detection, analysis, containment, eradication, recovery, lessons learned. NIST SP 800-61 Rev. 3 maps it to CSF 2.0.

Preparing to respond · Going deeper

The response team, playbooks and exercises

Preparation means an incident response plan, defined roles, playbooks and tabletop exercises.

Roles

  • Incident commander runs the response and makes the calls the plan assigns.
  • Technical lead and responders investigate, contain and eradicate.
  • Communications staff, customers and media, one voice.
  • Legal and privacy notification duties, evidence, law enforcement, the data protection officer.
  • Executives business decisions such as shutting down a service or engaging outside help.

Making it real

Planscope, severity levels, authority, contacts
Playbooksstep-by-step for common cases: ransomware, phishing, lost device
Tabletoptalk through a scenario around a table
Simulationrun it for real against test systems

Outside help is arranged in advance: an incident response retainer, cyber insurance and its approved firms, and law enforcement contacts.

Running an incident

The hardest part of a response is the decisions, made at speed

An example ransomware incident, from first alert to reviewMon 09:12DetectionEDR alert: thousands of file renames per minute on file server FS-0209:40Analysisconfirmed: ransomware encrypting shares; two servers affected so far09:55Containment◆ Decision: isolate the whole server segment now, accepting a four-hour outage10:30Containment◆ Decision: capture memory before anything is shut down or rebuilt11:00Analysis◆ Decision: legal and the DPO are told; personal data may be involved, so the 72-hour clock runsTueEradicationreset every admin password, remove persistence, patch the VPN flaw used to get inWed–ThuRecoveryrestore from offline backups, scan, return to service, monitor closely+ 2 weeksLessons learnedblameless review: what worked, what to change, an owner and date for each fixThe hardest part of a response is the decisions, and the best ones were made before the incident.Isolate now and accept an outage, or watch to learn more? Preserve evidence or restore fast? Who must be told,and by when? A response plan names who decides, and playbooks pre-agree the common answers, so the teamis not debating authority at 10 a.m. on the worst Monday of the year.

A real response is a sequence of decisions on containment, evidence and notification, made by the people the plan names.

Containment

Contain fast, but know what each action costs you

Containment actionStopsCostsEvidenceIsolate the host (EDR)spread from that machineone user offlinememory preservedDisable or reset accountsuse of stolen credentialsmay alert the attackerlogs preservedBlock IPs, domains, hashesknown command and controlattacker changes serversunaffectedCut off a network segmentspread between zonesa business outageunaffectedPull the powereverything on that machineall memory evidencelostIsolate rather than power off: memory holds the running malware, its connections and often its keys.Contain fast, but choose consciously what each action costs you in evidence and business.Short-term containment stops the spread now; long-term containment keeps systems running safely while thereal fix is built. Occasionally a team watches an attacker in a sandbox to learn more, but only by explicitdecision, because every minute of watching is a minute of risk.

Containment options trade speed against business impact and evidence; isolate rather than power off.

Digital forensics

Collect the evidence that disappears fastest, first

Collect evidence from the most volatile to the least1 CPU registers and cachenanoseconds2 Memory: processes, connections, keysuntil power-off3 Temporary files and swapminutes to hours4 Diskuntil overwritten5 Remote logs (SIEM)the retention period6 Backups and archivesyearsbar: how long it survives (not to scale)Evidence vanishes at different speeds, so collect the most fragile first.Memory disappears at power-off and changes every second; a disk lasts until overwritten; logs and backups lastmonths or years. RFC 3227 sets this order of volatility. Every action changes the system, so responders recordwhat they did and when, use trusted tools, and image disks through a write blocker.

The order of volatility (RFC 3227): registers, memory, temporary files, disk, remote logs, archives.

Forensic practice · Going deeper

Acquisition, imaging, timelines and legal hold

Know acquisition, write blockers, hashing for integrity, timelines, legal hold and e-discovery.

  • Acquisition capture memory with a trusted tool, then image disks bit for bit. In the cloud, snapshot volumes and export logs before instances are terminated.
  • Write blocker hardware or software that lets you read a drive without changing a single bit.
  • Hashing a SHA-256 of the image at collection and before every analysis proves integrity.
  • Timeline events from all sources merged on one clock (record time zones and clock drift) to tell what happened, in order.
  • Legal hold an instruction to preserve all potentially relevant data once litigation is reasonably expected, overriding normal deletion.
  • E-discovery finding and producing electronic records for legal proceedings.
Document everything: who did what, when, with which tool. An unexplained change to evidence is a gift to the other side’s lawyer.

Chain of custody

A chain of custody proves the evidence was never touched

ResponderTue 10:41images the disk, bags,tags and signssigned hand-offSHA-256 664033ca71…✓ hash recordedEvidence lockerTue 12:05logged in by thecustodian, sealedsigned hand-offSHA-256 664033ca71…✓ seal intactForensic analystWed 09:20works on a copy,never the originalsigned hand-offSHA-256 664033ca71…✓ hash matchesCourt or regulatormonths laterfull log of everyhand-off presentedSHA-256 664033ca71…✓ hash matchesEvery hand-off is logged: who, what, when, where and why. A gap can make the evidence inadmissible.A chain of custody proves the evidence in court is exactly what was collected, untouched.The image’s hash is recorded at collection and checked at every step; a match proves nothing changed. Analystswork on verified copies. A legal hold stops routine deletion of anything relevant once litigation is likely.

A chain of custody records every hand-off; matching hashes prove integrity; a legal hold preserves data.

Notification

Reporting deadlines are counted in hours, not weeks

How long after a breach you have to report it (log scale)NIS2: early warning to the CSIRT24 hoursGDPR: personal-data breach to the regulator72 hoursNIS2: incident notification72 hoursSEC 8-K: material incident (US listed)4 business daysNIS2: final report1 monthHIPAA: notice to affected individuals60 days1 h1 day3 days1 week1 month2 monthsThe clock usually starts when you become aware, not when the investigation ends.Reporting deadlines are measured in hours, so who calls whom is decided before the incident.Regulators, customers, insurers, law enforcement and partners may all need telling, each on their own clock andthrough set channels. Plans list the contacts and templates; legal counsel decides what is said. Inside theteam, use out-of-band communication if the attacker might be reading company email.

GDPR 72 hours; NIS2 24 hours, 72 hours and one month; SEC 8-K four business days; HIPAA 60 days.

Lessons learned

Lessons learned keeps asking “why?” until it reaches a fix

Ransomware encrypted two file serversWhy? The attacker had domain admin rightsWhy? Reused on a VPN account that was phishedWhy? The VPN accepted a password alone, with no MFAWhy? An old VPN exception was never reviewedFixes, each with an ownerMFA on every remote accessseparate admin accountsexceptions expire after 90 daysLessons learned asks “why?” until it reaches a cause you can fix, not a person to blame.Stopping at “an admin was phished” fixes nothing. Asking why repeatedly (the five whys) reaches the process gapthat let a phish become a disaster. Reviews are blameless, so people report honestly, and every action gets anowner and a date. Track mean time to detect and to respond across incidents to show whether it is working.

Root cause analysis, such as the five whys, in a blameless review; track MTTD and MTTR.

Lesson complete

You finished “Incident response”

Nicely done.

Open in a new tab